An MltA-Like Lytic Transglycosylase Secreted by Bdellovibrio bacteriovorus Cleaves the Prey Septum during Predatory Invasion

ABSTRACT Lytic transglycosylases cut peptidoglycan backbones, facilitating a variety of functions within bacteria, including cell division, pathogenesis, and insertion of macromolecular machinery into the cell envelope. Here, we identify a novel role of a secreted lytic transglycosylase associated with the predatory lifestyle of Bdellovibrio bacteriovorus strain HD100. During wild-type B. bacteriovorus prey invasion, the predator rounds up rod-shaped prey into spherical prey bdelloplasts, forming a spacious niche within which the predator grows. Deleting the MltA-like lytic transglycosylase Bd3285 still permitted predation but resulted in three different, invaded prey cell shapes: spheres, rods, and “dumbbells.” Amino acid D321 within the catalytic C-terminal 3D domain of Bd3285 was essential for wild-type complementation. Microscopic analyses revealed that dumbbell-shaped bdelloplasts are derived from Escherichia coli prey undergoing cell division at the moment of Δbd3285 predator invasion. Prelabeling of E. coli prey peptidoglycan prior to predation with the fluorescent D-amino acid HADA showed that the dumbbell bdelloplasts invaded by B. bacteriovorus Δbd3285 contained a septum. Fluorescently tagged Bd3285, expressed in E. coli, localized to the septum of dividing cells. Our data indicate that B. bacteriovorus secretes the lytic transglycosylase Bd3285 into the E. coli periplasm during prey invasion to cleave the septum of dividing prey, facilitating prey cell occupation. IMPORTANCE Antimicrobial resistance is a serious and rapidly growing threat to global health. Bdellovibrio bacteriovorus can prey upon an extensive range of Gram-negative bacterial pathogens and thus has promising potential as a novel antibacterial therapeutic and is a source of antibacterial enzymes. Here, we elucidate the role of a unique secreted lytic transglycosylase from B. bacteriovorus which acts on the septal peptidoglycan of its prey. This improves our understanding of mechanisms that underpin bacterial predation.

MltA Ec , Bd3285 contains a predicted lipoprotein signal peptide (aa 1 to 20) and a C-terminal 3D (three aspartate) domain (aa 271 to 326) (Fig. 1B). The three important aspartate residues of the MltA Ec catalytic 3D domain are conserved in Bd3285, including residue D321 (MltA Ec D328 ) which is crucial for the catalytic function of MltA Ec (31) (Fig. 1B; Fig. S2). Unique to Bd3285, and following the lipoprotein signal peptide, is a long, disordered Nterminus (aa 26 to 169) (Fig. S3). This region is absent from MltA Ec and also from Bd0519 and Bd0599 which solely encode a 3D domain, each containing the three conserved aspartate residues ( Fig. S1; Fig. S2). (1) Attack-phase (AP) B. bacteriovorus cells swim or glide to locate Gram-negative prey bacteria to which they attach (2) and then physically invade, concurrently rounding the rod-shaped prey cell into a spherical prey "bdelloplast" (3). B. bacteriovorus elongates as a filament within the prey periplasm, consuming the nutrients of the dead prey (4) until nutrients are exhausted and the predator divides to yield multiple progeny cells (5). B. bacteriovorus progeny then lyse the dead prey cell and seek out new prey to invade (6). (B) Genetic locus of the monocistronic bd3285 gene within the genome of B. bacteriovorus HD100 (top) and a schematic of the predicted domains and regions of the Bd3285 protein (bottom). Lipo-SP, lipoprotein signal peptide; 3D domain, domain containing 3 aspartate residues (D275, D309, and D321) predicted to be critical for catalysis based on homology to E. coli K-12 MG1655 MltA. (C) Reverse transcriptase PCR performed on B. bacteriovorus HD100 RNA isolated at time points during a predatory cycle on E. coli S17-1 prey. dnaK is a constitutively transcribed control gene. L, molecular weight 100 bp ladder; AP, attack-phase predators; 0.25 to 4, hours since predation commenced; NT, no template RNase-free water; Ec, E. coli S17-1 RNA, G, B. bacteriovorus HD100 genomic DNA. Data are representative of three biological repeats.
Upregulation of B. bacteriovorus gene expression at a particular time point during predation is often indicative of a particular predatory function; the DD-endopeptidase genes bd0816 and bd3459 are both upregulated at 0.25 to 0.5 h, consistent with their role in sculpting rod-shaped prey into spherical bdelloplasts during prey invasion (29) (Fig. 1A). Reverse-transcriptase PCR with mltA Bd -specific primers on RNA isolated from different time points during a predatory cycle on E. coli S17-1 prey showed that all three B. bacteriovorus mltA-like genes were upregulated 0.25 to 0.5 h into the predatory life cycle (Fig. 1C for bd3285; Fig. S4 for bd0519; Fig. S5 for bd0599), perhaps suggesting a role for these proteins in prey invasion processes.
Bd3285 is important for the transformation of prey shape during invasion. To determine whether the B. bacteriovorus MltA-like LTs are involved in prey invasion, markerless deletions of each gene were constructed and verified within the genome of B. bacteriovorus HD100. To visualize B. bacteriovorus predators inside prey bdelloplasts during predation, a fluorescent fusion of Bd0064-mCherry (a cytoplasmic PilZ domaincontaining protein, present throughout predation) was introduced into the genomes of B. bacteriovorus wild-type and deletion mutant strains by single-crossover homologous recombination. Fusions of fluorophores to the C-terminus of Bd0064 have been used extensively in previous studies to successfully label the cytoplasm of predator cells without adverse effects (28,35,36).
Wild-type and deletion mutant predator strains were mixed with stationary-phase E. coli S17-1 prey to initiate a predatory time course and images were acquired at different time points. Notably, in contrast to E. coli prey invaded by wild-type predators which all rounded up into spherical bdelloplasts, E. coli prey invaded by the Dbd3285 mutant formed three distinct shapes: spheres, rods, and rods with a midcell constriction (hereafter referred to as "dumbbells") ( Fig. 2A). While rod-shaped bdelloplasts have been observed before during predation on E. coli prey with a DD-endopeptidase Dbd0816/Dbd3459 predator mutant (29), the dumbbell shapes observed in predation with the Dbd3285 mutant were unique.
Complementation of the Dbd3285 predator mutant by double-crossover homologous reintegration of the wild-type bd3285 gene (WT comp) completely restored wild-type spherical bdelloplast morphology ( Fig. 2C and D). In contrast, complementation of Dbd3285 with a copy of bd3285 containing a mutation of the predicted catalytic aspartate residue D321 (D321A comp), failed to restore wild-type spherical bdelloplast shape, indicating that the 3D domain residue D321 is critical for Bd3285 function ( Fig. 2C and D). Stability of the Bd3285 (D321A comp) allele was verified by Western blotting analysis of an mCherry-tagged version with the wild-type Bd3285-mCherry (Fig. S7).
Deletion of bd0519 resulted in delayed prey cell entry with only 68.9% 6 SD 8.0% of prey cells containing a B. bacteriovorus predator 30 min after predator-prey mixing compared to 92.7% 6 SD 3.0% for the wild-type (Fig. S9). There was no significant defect in prey entry for Dbd0599 or Dbd3285 mutant predators, with 94.6% 6 SD 2.0% and 95.5% 6 SD 2.5% of prey cells invaded at 30 min, respectively (Fig. S9). The delay in prey entry for the Dbd0519 mutant resulted in a slightly prolonged predatory cycle with more bdelloplasts still visible 5 h after predator-prey mixing in comparison to the wild-type (Fig. S9). Thus, both other MltA Bd proteins have evolved different functions in the predatory life cycle, with the function of Bd0599 remaining unclear and Bd0519 Dumbbell-shaped bdelloplasts are derived from dividing E. coli prey. As the shape of dumbbell bdelloplasts from Dbd3285 predation resembled that of dividing E. coli, we hypothesized that dumbbell bdelloplasts originate from E. coli cells that were undergoing cell division at the moment of invasion by B. bacteriovorus. Invasion by Dbd3285 (concurrent with the death of the dividing prey cell) would result in persistence of the septating E. coli shape, forming a dumbbell bdelloplast.
In support of this hypothesis, there was no significant difference (P = 0.10) between the proportion of E. coli cells dividing prior to predation (6.8% 6 SD 1.1%) and the proportion of dumbbell bdelloplasts formed during predation by B. bacteriovorus Dbd3285 (4.3% 6 SD 1.8%) (Fig. S10). Time-lapse microscopy visualized individual predatory invasion events and, as predicted, wild-type predators sculpted rod-shaped E. coli cells into spheres during invasion (Fig. 3A). During invasion by Dbd3285 predators, rod-shaped prey were either sculpted into spheres or shortened in length but remained rod-shaped. Dividing E. coli prey that were invaded by Dbd3285 predators shortened slightly with some rounding at the cell poles, and the fixed septum producing a dumbbell shape (Fig.  3B).
These data suggest that the unique dumbbell-shaped bdelloplasts observed in the Dbd3285 mutant are formed from E. coli prey cells that were undergoing cell division until invasion by B. bacteriovorus, resulting in death of the prey and fixation of dividing cell shape.
Bd3285 localizes to and cleaves the septum of dividing E. coli prey. Considering that dividing prey could not be sculpted into spherical bdelloplasts by the Dbd3285 mutant, we hypothesized that Bd3285's wild-type role may be to interact with the peptidoglycan at the septum of dividing prey bacteria. Since Bd3285 contains an N-terminal lipobox-containing signal peptide, it is likely that the protein is secreted from B. bacteriovorus predator cells into the prey periplasm allowing access to the prey septum. Similar secretion patterns from predator into prey have been seen for two other lipobox proteins that modify prey-PG walls: novel lysozyme Bd0314 (DslA) which recognizes deacetylated peptidoglycan (27), and one of the prey-PG GlcNAc deacetylating enzymes Bd3279 (25). We tested this by constructing a fusion of Bd3285-mCherry within the chromosome of B. bacteriovorus in combination with a fusion of Bd0064-mCerulean3 (to label the cytoplasm of predator cells within prey bdelloplasts). During predation of this dually labeled B. bacteriovorus strain on E. coli, we observed that Bd3285-mCherry (while showing very faint signal, which is common for low level, high activity, B. bacteriovorus PG-active enzymes [26]) localized separately from the predator cell within the prey bdelloplast, suggesting that Bd3285 is indeed secreted into prey (Fig. S11). The Bd3285-mCherry fusion also retained functional activity as determined by complementation of the Dbd3285 mutant (Fig. S12).
We also tested the cellular location of Bd3285-mCherry protein heterologously expressed in E. coli TOP10 cells under the control of an arabinose-inducible pBAD vector promoter. Bd3285-mCherry showed heterogeneity in fluorescence brightness, with individual cells varying from no fluorescence to extremely bright fluorescence (Fig. S13). Some cells appeared visibly damaged by overexpression of Bd3285-mCherry which may account for the variation in fluorescence intensity (Fig. S13) but heterologous expression did give an opportunity to examine cellular destinations of the protein. In E. coli TOP10 cells exhibiting optimal brightness, Bd3285-mCherry protein appeared to localize most strongly to the periplasmic compartment ( Fig. 4A and B). Most interestingly, in cells that had a constriction at the midcell and were undergoing division, Bd3285-mCherry primarily localized to the midcell septum ( Fig. 4B;  Fig. S14). It is important to note that the shape of these dividing E. coli TOP10 cells expressing just Bd3285 alone differs from the dumbbell bdelloplasts observed during predation by the Dbd3285 B. bacteriovorus mutant (the cell compartments are less rounded). This is because in E. coli TOP10 (Fig. 4), only the Bd3285 protein is being expressed, whereas during predation experiments, a full consignment of predator enzymes are secreted into E. coli prey, including two DD-endopeptidase enzymes which contribute to the sculpting of a rounded prey bdelloplast shape (discussed later in further detail).
As Bd3285 localizes to the E. coli septum and the protein is a predicted lytic transglycosylase, we hypothesized that Bd3285 may cleave the septum of dividing prey cells. If true, then the septum should still be intact within dumbbell-shaped bdelloplasts. To test this hypothesis, we used the fluorescent D-amino acid (FDAA) HADA to prelabel the E. coli PG cell wall prior to predation. B. bacteriovorus wild-type and Dbd3285 strains (each containing a Bd0064-mCherry fusion to label the cytoplasm) were incubated with prelabeled E. coli S17-1 for 30 min and then samples were removed for imaging. E. coli bdelloplasts formed by invading wild-type B. bacteriovorus were spherical in shape and blue HADA signal was observed throughout the bdelloplast sphere, with HADA uniformly labeling all PG (Fig. 5A). Spherical and rodshaped bdelloplasts formed by Dbd3285 predators also showed uniform HADA incorporation (Fig. 5A). Critically, all dumbbell-shaped bdelloplasts showed additional fluorescent HADA signal across the midcell, labeling septal PG ( Fig. 5B; Fig.  S15). Interestingly, some rod-shaped bdelloplasts had HADA signal at the midcell sidewalls which may represent preseptal PG as the prey cell prepares to undergo cell division (Fig. 5C). This preseptal PG might prevent full conversion of these rods into spheres by other predatory enzymes in the absence of Bd3285 (discussed in further detail later). The complementation strain Dbd3285 (WT comp) formed solely spherical bdelloplasts with a uniform HADA signal (Fig. S16A), while predation with the Dbd3285 (D321A comp) strain resulted in a mixture of spheres, rods, and dumbbell bdelloplasts (with HADA labeling of the dumbbell septum) (Fig. S16A and B).
Collectively, these data show that the PG septum of dividing E. coli prey cells remains, and is not cleaved upon invasion by Dbd3285 predators, indicating that B. bacteriovorus lytic transglycosylase Bd3285 is involved in the cleavage of prey septal PG. Failure to cleave the septum results in prey bdelloplasts that are not sculpted into optimal bdelloplast spheres.

DISCUSSION
During invasion into Gram-negative prey, the bacterial predator B. bacteriovorus secretes an arsenal of PG-modifying enzymes to modify the prey cell wall, creating an entry porthole through which the predator can access the prey periplasmic compartment. Here, we identify a novel lytic transglycosylase lipoprotein, Bd3285, which is secreted by B. bacteriovorus into E. coli prey to cleave the PG septum of prey cells undergoing cell division.
Three B. bacteriovorus proteins, Bd3285, Bd0599, and Bd0519, share homology with the lytic transglycosylase MltA from E. coli K-12 MG1655, including strong conservation of a C-terminal catalytic 3D domain (Fig. S1). Although MltA Ec has been well characterized structurally and biochemically (31, 37), a role for MltA Ec in vivo has not been demonstrated (32). We discovered that all three B. bacteriovorus mltA-like genes were upregulated at the time point of prey invasion ( Fig. 1C; Fig. S4 and 5). While no phenotype could be determined for Bd0599, prey invasion was delayed in a Dbd0519 mutant, but not in the Dbd3285 mutant (Fig. S9).
Most striking, however, was the observation that a proportion of prey cells invaded during predation by a Dbd3285 mutant did not round up into spherical bdelloplasts, in contrast to invasion by wild-type B. bacteriovorus (Fig. 2) and in contrast to Dbd0519 Dbd0599 mutants (Fig. S8). Additional experiments revealed that Bd3285 is secreted from the indwelling predator into the prey periplasm (Fig. S11), can localize to a prey septum (Fig. 4), and is required to cleave the septum of these dividing prey during bdelloplast formation (Fig. 5).
The phenotypic diversity of these similar MltA-like LT proteins contrasts with that found for self-LT processes in V. cholerae (10). There, the absence of a specific LT was compensated for by expression of another heterologous LT (10). In B. bacteriovorus, neither the lipobox-containing MltA homologue Bd0599 nor the sec-signal-containing Bd0519 compensate for Bd3285 in prey septal-processing.
Bd3285 secreted by B. bacteriovorus appears to preferentially localize to the prey septum and this is a protein-intrinsic property also seen when Bd3285 is heterologously expressed in E. coli. Septal localization has also been observed for the LT proteins RlpA (in both Pseudomonas aeruginosa [5] and Vibrio cholerae [6]) and MltC in Vibrio cholerae (6). Both proteins are involved in daughter cell separation (5,6). RlpA contains a SPOR domain which specifically binds septal PG by preferentially recognizing denuded glycan chains at the septum (5,38,39). However, Bd3285 does not contain a SPOR domain nor any other domains known to bind differentially modified PG, suggesting that its mode of recruitment and binding to PG may differ from other proteins. It would be interesting to test the catalytic properties of Bd3285 in vitro; however, we were unable to produce soluble Bd3285, having been guided by an Alphafold model to clone region S204-K420 into pET29b. Trial expression in BL21 cells yielded a clear signal which unfortunately was entirely in the insoluble fraction and was not amenable to standard refolding protocols.
Bd3285 has a lipobox motif and a disordered N-terminal domain which may be involved in a novel secretion and targeting mechanism. This must initially target Bd3285 to the B. bacteriovorus periplasm, then outer membrane, and from there it may be inserted into either the prey inner or outer membrane, or the lipid anchor may be cleaved to release Bd3285 into the prey periplasm. This may then diffuse to interact with the prey PG, preferentially targeting the septum. An alternative is that Bd3285 may be packaged into outer membrane vesicles that are delivered to the prey; such vesicles have recently been observed in bdelloplasts via cryo-electron tomography (40). As B. bacteriovorus has a wide prey range, this novel targeting would have to be of a general nature rather than by specific interactions with other proteins as these would vary considerably between different prey. It is noteworthy that the timing of expression of Bd3285 and its action on the prey cell septum comes early in the predatory cycle before the predator itself is forming septa.
Previous work by Lerner et al. uncovered the role of two DD-endopeptidase (DacB) enzymes which are secreted by B. bacteriovorus during prey invasion (29). Deletion of both dacB genes resulted in a population of ;95% rod-shaped bdelloplasts (29). Comparing this to the morphology of bdelloplasts produced by Dbd3285 mutant predators where we found that 33.0% formed rods, we suggest that it is likely that Bd3285 has evolved to act in concert with the DacB enzymes to optimize prey invasion by wild-type predators. During predation by Dbd3285 predators, DacB activity alone may be sufficient to sculpt most short rod-shaped E. coli cells into spheres as 62.7% of bdelloplasts from the Dbd3285 mutant are spherical. For the ;33.0% of rod-shaped prey bdelloplasts from the Dbd3285 mutant that shorten in length but remain rod-shaped, it is possible that a ring of preseptal PG (not cut by Bd3285 in Dbd3285) may hold the rod shape in place, preventing complete cell rounding by other enzymes during predation. Equally, for prey cells undergoing a later stage of cell division at the point of predator invasion, septal PG (not cut by Bd3285 in Dbd3285) could also fix this constricted septating cell shape in place, resulting in dumbbell-shaped bdelloplasts. In both of the latter scenarios, it is possible that the DacB enzymes, Bd0816 and Bd3459, can access PG cross-links at the midcell sidewall, but the presence of septal or preseptal PG (that is not cleaved by Bd3285) fixes the prey shape in place, preventing conversion into spherical bdelloplasts. During predation by wild-type B. bacteriovorus, cleavage of septal PG by Bd3285 would therefore be required to facilitate the complete conversion of E. coli rods and dividing cells by other PG-active enzymes (particularly DacBs) into spherical bdelloplasts (Fig. 6). Such a uniformly spherical shape may distribute mechanical stresses, from the predator within, evenly across the bdelloplast envelope, preventing premature lysis prior to predator replication.
Considering the evolution of predation, cleavage of the septum of a prey cell to form a fully rounded bdelloplast likely creates an optimized niche for predator growth, enabling access to the whole prey contents and room for filamentous growth throughout the bdelloplast. Therefore, we suggest that, in addition to generating strong, stable spherical bdelloplasts, another predatory population-fitness advantage conferred by Bd3285 septum-cleavage may be to provide a single invading B. bacteriovorus predator cell access across any prey septa to maximal prey nutrients during intra-bdelloplast growth without a second invasion being required on the distal side of the septum. Unfortunately, we cannot quantify any fitness change experimentally due to the varying and low percentage of septating prey, but we suggest that evolution of Bd3285 may have arisen as an additional "efficiency benefit" to basic predation in this way.
In summary, we have uncovered a novel function for a lytic transglycosylase lipoprotein which is secreted by B. bacteriovorus into prey bacteria to cleave the septal PG of dividing cells, allowing each predator access to the whole prey as a niche, whether a septum is present or not. These findings enhance our fundamental understanding of the life cycle of a bacterial predator that has potential application in future antibacterial therapies.

MATERIALS AND METHODS
Bacterial culture. Bdellovibrio bacteriovorus strain HD100 (Type strain) was routinely cultured in liquid calcium/HEPES buffer (5.94 g/L HEPES free acid, 0.284 g/L calcium chloride dihydrate, pH 7.6) containing Escherichia coli S17-1 (;1 Â 10 9 CFU/mL) in a predator to prey ratio of 1:3 as described previously (41). For growth on solid media, B. bacteriovorus was cultured on double-layer YPSC plates as described previously (41). Kanamycin-resistant E. coli S17-1 (pZMR100) was substituted as prey for the culture of kanamycin-resistant B. bacteriovorus strains. E. coli strains used as bacterial prey were routinely cultured on either solid YT or in liquid YT media, inoculated by a single colony and incubated at 37°C for 16 h, with orbital shaking at 200 rpm. Media were supplemented with kanamycin (25 mg/mL) when required.
Reverse-transcriptase PCR. To measure the transcriptional pattern of mltA-family genes, versus the dnaK constitutively expressed control, reverse transcriptase PCR (RT-PCR) was carried out on RNA template isolated from different time points during predation of B. bacteriovorus HD100 on E. coli S17-1 using an SV Total RNA Isolation System kit (Promega) as described previously (42). The quality of RNA template was verified on an Agilent Bioanalyzer using the Agilent RNA 6000 Nano Kit. RT-PCR was performed using the Qiagen OneStep RT-PCR kit using the parameters: 50°C for 30 min, 94°C for 15 min, bacteriovorus upon E. coli prey that is either (i) non-dividing or (ii) dividing at the moment of predator invasion. Non-dividing prey does not contain any septal peptidoglycan; therefore, de-cross-linking activity of DacB enzymes alone is sufficient to convert rod-shaped prey into spherical bdelloplasts (i). In dividing prey with septal peptidoglycan, however, Bd3285 lytic transglycosylase action is required to cleave the septum, facilitating full conversion of dividing cells into spherical bdelloplasts (ii). (B) Close-up model of Bd3285 action at the septum during predation on dividing prey cells.
Prey Septum Cleavage by B. bacteriovorus Novel MltA Journal of Bacteriology followed by 30 cycles of 94°C for 1 min, 50°C for 1 min, and 72°C for 1 min, and a final step of 72°C for 10 min. Samples were visualized on a 2% agarose gel run at 100 V for 20 min. Construction of deletion and complementation strains. Primers and plasmids used to construct strains used in this study are detailed in Table S1 and Table S2. Bacterial strains are listed in Table S3. Bd3285 amino acid numbering was updated to correct for a mis-annotated start codon (the true protein starts at M44; Fig. S14).
To construct in-frame silent genetic deletions in B. bacteriovorus, 0.5 to 1 kb of DNA flanking the gene of interest was cloned into the suicide vector pK18mobsacB by Gibson Assembly (43) using the NEBuilder HiFi DNA Assembly Cloning Kit (New England Biolabs). Genetic constructs were transformed into the donor strain E. coli S17-1 and conjugated into B. bacteriovorus as described previously (29,44). Double-crossover deletion mutant exconjugants were generated by sucrose suicide counter-selection and verified by Sanger sequencing. To test complementation of Dbd3285, either a wild-type copy of bd3285 or a copy of bd3285 containing the point mutation D321A (catalytic domain mutant) was introduced into the Dbd3285 mutant by double crossover homologous recombination as described above and verified by Sanger sequencing. The Dbd3285 mutant was also complemented with bd3285-mCherry in trans on the vector pMQBAD, derived from pMQ414 (45), by cloning the bd3285 gene plus 100 bp of flanking DNA into the vector which was then conjugated into B. bacteriovorus Dbd3285.
Construction of fluorescent fusions and overexpression strains. To label the cytoplasm of B. bacteriovorus and thus allow visualization of predator cells within prey bdelloplasts, a fluorescent fusion of Bd0064-mCherry was introduced into B. bacteriovorus strains via single-crossover homologous recombination and maintained under kanamycin selection.
To construct a single-crossover fluorescent fusion of Bd3285-mCherry, the gene (minus the signal peptide and stop codon and with mCherry fused to the 39 end) was assembled into pK18mobsacB using Gibson Assembly. Constructs were conjugated into B. bacteriovorus containing a double-crossover fluorescent fusion of Bd0064-mCerulean3, verified by Sanger sequencing and maintained under kanamycin selection.
To construct a strain of E. coli overexpressing Bd3285-mCherry, the bd3285 gene (minus the stop codon and with mCherry fused to the 39 end) was cloned into the vector pBAD HisA under the control of an arabinose-inducible araBAD promoter. The construct was transformed into E. coli TOP10, verified by Sanger sequencing and maintained under kanamycin selection.
Phase contrast and epifluorescence microscopy. Cells were pipetted onto a thin 1% Ca/HEPES agarose pad and imaged under the Plan Apo Â100 Ph3 oil objective lens (NA: 1.45) of an inverted Nikon Ti-E epifluorescence microscope. The following filters were used to acquire fluorescence images: mCherry (excitation: 555 nm, emission: 620/60 nm), DAPI: (excitation: 395 nm, emission: 435 to 485 nm), and mCerulean3 (excitation: 440 nm, emission: 470 to 490 nm). Images were captured on an Andor Neo sCMOS camera using Nikon NIS software. To overexpress Bd3285, E. coli TOP10 cells were either induced with 0.2% L-arabinose for 20 h at 37°C. Samples were then removed and placed onto a 1% agarose pad for imaging.
Labeling of the peptidoglycan cell wall. To visualize the prey cell wall during predation, E. coli prey were prelabeled with the fluorescent D-amino acid (FDAA) cell wall stain HADA (46) (kind gift from Dr. Erkin Kuru and Prof. Michael VanNieuwenhze, Indiana University) prior to predation. In brief, E. coli S17-1 was cultured for 16 h, adjusted to an OD 600 = 1.0 in fresh YT medium, mixed with 500 mM HADA, and incubated at 29°C for 30 min. Cells were then microcentrifuged at 17,000 g for 5 min and the pellet washed with Ca/HEPES. The process was repeated to remove any unincorporated free HADA stain.
Predatory cultures for a semisynchronous time course were set up in a 5:4:3 ratio of 10-fold concentrated predator cells: HADA-labeled E. coli: Ca/HEPES buffer. After 30 min, 120 mL was removed and transferred to a microcentrifuge tube containing 175 mL of precooled 100% ethanol. Tube contents were mixed by inversion, incubated at 220°C for 15 min, then microcentrifuged at 17,000 g for 5 min and the pellet resuspended in 500 mL of 1Â PBS. Samples were microcentrifuged again and the pellet resuspended in 5 mL of Slow-Fade (Molecular Dimensions) and stored at 220°C. For imaging, 2 mL of each sample was transferred to a microscope slide and imaged as a wet mount. HADA fluorescence was captured with the DAPI filter and a 1 s exposure time, while B. bacteriovorus Bd0064-mCherry fluorescence was captured with the mCherry filter and a 10 s exposure time.
Time-lapse microscopy. Time-lapse microscopy videos of B. bacteriovorus invasion into E. coli prey cells were captured under the Â100 oil objective lens (NA: 1.25) of an upright Nikon Eclipse E600 microscope. To prepare samples, 1 mL of attack-phase B. bacteriovorus and 50 mL of stationary-phase E. coli S17-1 were separately microcentrifuged at 17,000 g for 2 min and resuspended in 50 mL of Ca/HEPES. Predators and prey were mixed and transferred onto a soft 0.3% Ca/HEPES agarose pad. Time-lapse images of multiple fields of view were captured every 1 min for at least 2 h using a motorized Prior Scientific H101A XYZ stage and Hammamatsu Orca ER Camera with Simple PCR software.
Image analysis. The Fiji distribution of ImageJ (47) was used for image processing and analysis. Images were sharpened and smoothed and minimal adjustments were made to brightness and contrast. The MicrobeJ plug-in for Fiji (48) (v. 5.11z) was used to detect B. bacteriovorus (labeled with Bd0064-mCherry) within prey bdelloplasts and measure the different morphologies of bdelloplasts. B. bacteriovorus predators within bdelloplasts were detected in the Maxima tab using the "Bacteria" setting. Bdelloplasts were generally identified by the parameters of area: 1.0-max mm 2 , length: 0.5-max mm, width: 0.5-max mm, curvature 0 to 0.35 A.U., and circularity: 0.6 to 1.0 A.U. Circularity is defined by MicrobeJ as "4p r x area/perimeter 2 , with a value of 1.0 indicating a perfect circle." Bdelloplasts with a circularity value of #0.96 A.U. were classified as nonspherical based on visual observations and the fact that approximately 0% of wild-type bdelloplasts had a circularity score of #0.96 A.U. The classification of nonspherical bdelloplasts allowed the proportion of spherical bdelloplasts to be quantified. All images were manually inspected to ensure that cells had been correctly detected. Bdelloplasts that either did not contain a B. bacteriovorus predator or had B. bacteriovorus cells attached to the outside (distorting shape measurements) were removed prior to analysis.
Western blot. Semisynchronous predation was set up for Bd3285 (WT)-mCherry and Bd3285 D312A-mCherry strains, as described above and at 30 min postmixing of predators and prey, 500 mL samples were concentrated to 100 mL by centrifugation at 17,000 Â g for 2 min. This was followed by the addition of 40 mL of 4Â loading buffer (2 mL 10% SDS, 0.5 mL 0.5% bromophenol blue, 600 mL 1M Tris pH 6.8, 350 mL water, 1.25 mL 80% glycerol, 500 mL b-mercaptoethanol) and samples were frozen at 220°C. Equivalent volumes of attack-phase B. bacteriovorus only and E. coli only controls were also collected and concentrated to 100 mL. Samples were boiled at 105°C for 5 min then 10 mL was loaded onto 4% to 20% SDS-PAGE gels with 3 mL MagicMark XP ladder (Invitrogen) or 10 mL SeeBlue Plus2 ladder (Invitrogen) for loading control gels, which were stained with QuickBlue Protein Stain (LubioScience). Gels were transferred onto a nitrocellulose membrane for 2 h at 25 V. The WesternBreeze Chemiluminescence kit (Novex) was used for immunodetection according to the manufacturer's instructions with anti-mCherry primary antibody (Invitrogen, product no: PA5-34974, diluted 1:2,000) incubated overnight at 4°C. Images were captured via exposure to X-ray film.
Statistical analysis. All statistical analyses were performed in GraphPad Prism 8.0. Collected data were first tested for normality to assess whether the data exhibited a Gaussian or non-Gaussian distribution. Data sets were subsequently subjected to the most appropriate statistical test. The statistical tests applied to data sets, P-values, and number of biological repeats performed for each experiment are described in the figure legends.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, DOCX file, 6.9 MB.

ACKNOWLEDGMENTS
The project was funded by a Wellcome Trust PhD studentship (215025/Z/18/Z) which supported E.J.B. and a Wellcome Trust Investigator Award in Science (209437/Z/ 17/Z) to R.E.S. and A.L.L., which supported C.L., J.T., P.M.R., A.L.L., and R.E.S., S.S.M and C.M were undergraduate students at the University of Nottingham. We thank Simon Caulton for trial Bd3285 protein expression.